Today, server utilization levels have increased and the number of servers required to do the same work has decreased, contributing to lower power and cooling costs and better energy efficiency.

In addition to server consolidation, organizations have also found that virtualization solves an even broader set of challenges. Disaster recovery solutions with virtualization are simple, reliable, and more cost effective. Capital and operating costs can be lowered in virtual desktop environments with the use of centralized servers and thin clients. Development, test, and production environments can coexist on the same servers, and application deployment can be decoupled from server purchasing decisions. With virtualization, new applications can be deployed and scaled on demand to accommodate changing business needs.

Increased Demand for Bandwidth

The combination of virtualization and multicore processors has been increasing the amount of work that each server can accomplish. But as sockets, cores, and memory slots on each server have been growing by powers of two, general server network connectivity has increased only a gigabit at a time.

With the increased server capacity comes an ever greater demand for I/O capacity, making network bandwidth a potential problem. But adding a gigabit network interface card (NIC) here and there can become costly because every new NIC increases server power consumption, cabling cost and complexity, and creates a greater number of access-layer switch ports that need to be purchased and managed.

And, to accommodate the number of needed expansion slots, companies must often purchase larger, more costly servers which can lower the ROI in virtualization and create a higher TCO.

Too Many Ports; Not Enough Bandwidth

A typical server running VMware ESX Server uses between four to eight Gigabit Ethernet links, creating an increased need for more NICs, upstream switch ports, and cabling. It is even common for a server to be configured with up to eight NICs using discrete ports for each function. But taking into consideration today’s workloads, even a 6-port configuration may be insufficient.

Making the Move to 10 Gigabit Ethernet

One way to tackle this challenge is to consider 10 Gigabit Ethernet which delivers 10 times the bandwidth, with some switches and NICs offering features that support multiple physical link consolidation, including the ability to create lossless priority classes for iSCSI, Network File System (NFS), and Fibre Channel over Ethernet (FCoE) traffic.

FCoE is supported on some 10 Gigabit Ethernet NICs as well as by a new class of converged network adapters that appear to the OS as separate Fibre Channel host bus adapters (HBAs) and Ethernet NICs, which makes FCoE’s existence entirely transparent to the OS.

So instead of using four to eight Gigabit Ethernet NICs in each server, organizations can deploy just two 10 Gigabit Ethernet NICs and reach full redundancy for availability plus additional expansion room. Some Gigabit Ethernet NICs support FCoE, and all support iSCSI and NFS traffic to access virtual disks.

Benefits of 10 Gigabit Ethernet

• Lower Cost


• Greater Overall Cost Effectiveness
  Due to the increase in deployment, the cost of 10 Gigabit NICs and switches has been decreasing, making it cost effective to configure two 10 Gigabit Ethernet NICs rather than 4 to 8 Gigabit Ethernet NICs.


• Fewer Access-layer Switches
  When you reduce the number of Gigabit Ethernet ports, you reduce the number of NICs, upstream switch ports, and cables you need, simplifying your data center network infrastructure.


• Lower Power and Cooling Costs
  Lowering the number of NICs and switch ports decreases a data center’s power and cooling load, increasing data center efficiency.


• Increased Scalability


• Increased Bandwidth Per Virtual Machine
  Because 10 Gigabit Ethernet allows each virtual machine to use bandwidth of more than 1 Gbps, performance is increased and network bottlenecks reduced. Many, if not most, organizations use active-active virtual port ID–based NIC teaming which limits virtual machines to 1 Gbps. But with 10 Gigabit Ethernet, each virtual machine can use more than 1 Gbps of bandwidth regardless of the type of virtual NIC teaming used.


• Support for Networked Storage
  Sophisticated features such as VMware HA and VMotion require the use of networked storage so that virtual disks can be accessed from any server. 10 Gigabit Ethernet offers greater bandwidth, lower latency, and easier management for iSCSI and NFS-based storage. With FCoE-enabled network adapters, 10 Gigabit Ethernet’s ability to carry Fibre Channel over Ethernet traffic makes it a drop-in replacement for traditional Fibre Channel SANs at the access layer. As with per–virtual machine production bandwidth, 10 Gigabit Ethernet eliminates the 1 Gbps bandwidth bottleneck for storage traffic.


• Simplified Management


• Ease of Management
  Consolidating a number of Gigabit Ethernet links onto a single 10 Gigabit Ethernet connection dramatically decreases the number of management points, including NICs and their firmware, switch ports, and cables.


• Simplified VMware VMotion Support
  Moving virtual machines from one server to another requires identical I/O configurations between servers. Standardizing on only one pair of 10 Gigabit Ethernet NICs per server helps ensure consistent I/O configurations on servers running VMware ESX Server.


• Fewer Points of Failure
  With a decreased number of devices, the number of failure points in the access layer is reduced which contributes to higher availability.

Conclusion

In order for organizations to begin to fully realize the benefits of virtualization, they must look at their infrastructure with a new set of eyes. Data flows are now converging in the data center—changing how traditional networks have been designed. Continuing to design systems they way we have since the late 90s with the new technologies on the market will not allow us to recognize all of the value that virtualization has to offer.

The next transition that everyone will recognize shortly is the Unified Computing Environment. Unified Computing is the advancement toward the next generation data center that links all resources together in a common architecture to reduce the barrier to entry for data center virtualization. In other words, the compute and storage platform is architecturally “unified” with the network and the virtualization platform. 10GB is and the systems that support 10GB will be at the forefront of this new and exciting technology.

‹ ‹ ‹ Back to Newsletter